Research projects in progress + future anticipated research projects:
Analysis of “combination mechanisms of action”
Mechanisms of action are defined as direct upstream biochemical effects in a direct drug target. In textbook biochemistry, the concentration of the drug determines the inhibition of the target. Think about first year biochemistry and enzyme kinetics! But, drugs bind lots of proteins at different affinities and they alter downstream cellular function through direct and indirect mechanisms that can be non-linear and exhibit feedback. Moreover, target biochemistry itself can be non-linear within a single protein when distant sites on the same molecule influence each other via allostery. Thus, the biology of combinations is far more complex than the addition of 2 distinct biochemical inhibitors. These indirect effects are especially interesting and important to think about combination modes of action. Pooled genetic methods are uniquely suited to understand the complexity of combination biology across entire dose response surfaces because they create unbiased maps of the modifiers of small molecule response across entire cells and proteins. These maps are important gauges of cell function and they have been used in E. coli, yeast, and mammalian cells to classify single drug mechanisms at single concentrations, and occasionally combinations. Very little work has been done with drug combinations at multiple concentrations, we aim to change this by defining mechanisms of combination action that are spatial across a dose response surface
Dual-Switch Selection Gene Drives to Combat Drug Resistance
Despite our best efforts, drug resistance evolution–from prokaryotes to cancers– is one of the largest threats to public health. We propose that a novel paradigm to understand and fight drug resistance evolution is to use forward engineering design. Instead of simply describing the outcome of resistance evolution after it has occurred, we will use model-driven design to build dual-switch selection gene drives, and we will test their ability to control resistant populations which will enable a deeper understanding of biology and a shift in clinical paradigms. These gene drives use cell therapies to achieve local combinations of cytotoxic small molecules.
CRISPR Lossy Compression
A genetic knockout can be lethal to one cell and increase the growth rate of another. This context specificity confounds our understanding of genetics and it prevents reproducible genome engineering across mammalian cells. 2 large collections of pooled CRISPR screens across nearly 1000 mammalian cell lines offer the largest opportunity to understand cell specificity systematically. The prevailing explanation, synthetic lethality, occurs when a single mutation in a genetic background creates a unique genetic dependency. To identify the dominant explanations of cell type specificity in an unbiased manner, we used machine learning to systematically mine millions of omic and CRISPR features. By quantifying the prediction accuracy of our models, we found that most cell type specific phenotypes are predicted by the function of related genes that are genetically wild type, and not synthetic lethality. The implications of these models led to a totally unexpected application of our finding wherein we identified “lossy compression” sets of 100-300 genes where reduced CRISPR measurements are sufficient to produce genome-scale loss-of-function predictions across >18,000 genes. These lossy CRISPR sets are a functional genomic analogy to in silico lossy compression of data. In lossy compression, it is possible to reduce in vitro CRISPR libraries by orders of magnitude—with some information loss—when we remove redundant genes and not redundant sgRNAs. This makes previously impossible genome-scale experiments possible.